Lloyd Snowdon

It has been common practice to plot gasoline range parameters as absolute
concentrations within crude oils of either single compounds, sums of compound pairs, or the square root of the products of concentrations of compounds. If a sample of crude such as a condensate has a relatively high concentration of volatile components, then the concentration of all of the gasoline range components tends to be high. Conversely, heavy oil with very low concentrations of volatile compounds tends to have low concentrations of each of the compounds in the gasoline range (C5 - C8 ). By using absolute concentrations, much of the apparent correlation between and among various gasoline range compound parameters derives simply from the differential scaling through three or four orders of magnitude which results from the simple presence or absence of large amounts of the gasoline fraction within a given crude oil sample. This variance due to scaling overwhelms the variation in the concentration of these components that may be ascribed to different genetic signatures or different generation or alteration processes to which the samples may have been subjected. Statistically, this is referred to as a “spurious correlation” between parameters. It is caused by a third, possibly unknown, “lurking variable”. To prevent “spurious correlation” parameters must only be considered within groups or categories of samples which are similar with respect to third or additional properties.